Recursive filter for multirate systems with missing measurements and random attacks

Raquel Caballero-Águila, University of Jaén

Co-authors: M. Pilar Frías, University of Jaén; Antonia Oya, University of Jaén

Abstract: This paper addresses the least-squares linear filtering problem for discrete-time stochastic signals based on uncertain measurements which may either contain noisy information about the signal to be estimated or consist only of noise. This phenomenon, commonly referred to as missing measurements or uncertain observations, often arises due to sensor saturation, limited sensing capability, or temporary sensor failures, among other factors. The study focuses on a class of multirate systems, where the signal update rate differs from the sampling rate of the sensor collecting the measurements. Additionally, the system is considered to be vulnerable to potential Denial-of-Service (DoS) attacks, which cause random data losses during transmission. To mitigate this issue, the hold-input compensation technique is employed, replacing lost measurements with the most recently received one. A recursive filtering algorithm based on covariance information is proposed, offering high versatility and broad applicability, as it does not require prior identification of the signal evolution model.